Talking to Strangers

by Malcolm Gladwell

  “I don’t know why…too simplistic for us”: Fox Butterfield, “A Way to Get the Gunmen: Get the Guns,” New York Times, November 20, 1994, https://www.nytimes.com/1994/11/20/us/a-way-to-get-the-gunmen-get-the-guns.html.

  In 1991 the New York Times: Don Terry, “Kansas City Police Go After Own ‘Bad Boys,’” September 10, 1991, https://www.nytimes.com/1991/09/10/us/kansas-city-police-go-after-own-bad-boys.html.

  For the rise in North Carolina traffic stops in the early 2000s, see Deborah L. Weisel, “Racial and Ethnic Disparity in Traffic Stops in North Carolina, 2000–2001: Examining the Evidence,” North Carolina Association of Chiefs of Police, 2014, http://ncracialjustice.org/wp-content/uploads/2015/08/Dr.-Weisel-Report.compressed.pdf.

  One of Weisburd’s former students (in footnote): E. Macbeth and B. Ariel, “Place-based Statistical Versus Clinical Predictions of Crime Hot Spots and Harm Locations in Northern Ireland,” Justice Quarterly (August 2017): 22, http://dx.doi.org/10.1080/07418825.2017.1360379.

  Chapter Twelve: Sandra Bland

  “Dude, issue the…pulling her out?”: Nick Wing and Matt Ferner, “Here’s What Cops and Their Supporters Are Saying about the Sandra Bland Arrest Video,” HuffPost, July 22, 2015. https://www.huffingtonpost.com/entry/cops-sandra-bland-video_us_55afd6d3e4b07af29d57291d.

  “An employee of the Department…extreme provocation”: Texas Department of Public Safety General Manual, Chapter 5, Section 05.17.00, https://www.documentcloud.org/documents/3146604-DPSGeneralManual.html.

  TSA haystack searches: DHS Press Office, “DHS Releases 2014 Travel and Trade Statistics,” January 23, 2015, https://www.dhs.gov/news/2015/01/23/dhs-releases-2014-travel-and-trade-statistics, accessed March 2019.

  “go beyond the ticket” and other Remsberg quotes: Charles Remsberg, Tactics for Criminal Patrol: Vehicle Stops, Drug Discovery, and Officer Survival (Northbrook, Ill.: Calibre Press, 1995), pp. 27, 50, 68. Also from this source: “If you’re accused…the defendant’s case,” p. 70; “concealed interrogation” and “As you silently analyze…incriminating evidence,” p. 166; and “Too many cops…what the suspect does,” pp. 83–84.

  the driver was “stiff and nervous”: Heien v. North Carolina, 135 S. Ct. 534 (2014), https://www.leagle.com/decision/insco20141215960.

  When he approached the stopped car: Gary Webb, “DWB: Driving While Black,” Esquire 131, issue 4 (April 1999): 118–27. Webb’s article was really the first to document the growing use of Kansas City techniques. It is superb—and chilling. At one point he sits down with a Florida officer named Vogel who was a particularly aggressive proponent of proactive searches. Vogel was proud of his sixth sense in spotting potential criminals. Webb writes: Other indicators, [Vogel] said, are adornments like “earrings, nose rings, eyelid rings. Those are things that are common denominators with people who are involved with crimes. Tattoos would go along with that,” particularly tattoos of “marijuana leaves.” Bumper stickers also give him a feel for the soul of the driver. “Deadhead stickers are things that almost—the people in those kinds of vehicles are almost always associated with drugs.”

  Give me a break.

  a day from Brian Encinia’s career: Los Angeles Times Staff, “Citations by Trooper Brian Encinia,” Los Angeles Times, August 9, 2015, http://spreadsheets.latimes.com/citations-trooper-brian-encinia/.

  “I was checking…yes sir” (and all Encinia/Renfro Q&A quotes from Brian Encinia): Interview with Cleve Renfro (Texas Department of Public Safety Lieutenant), October 8, 2015. Audio obtained by KXAN-TV of Austin, https://www.kxan.com/news/investigations/trooper-fired-for-sandra-bland-arrest-my-safety-was-in-jeopardy/1052813612, accessed April 2019.

  “An operator shall use the signal…”: Texas Transportation Code, Title 7: Vehicles and Traffic, Subtitle C: Rules of the Road, Chapter 545: Operation and Movement of Vehicles, Sections 104, 105, p. 16, https://statutes.capitol.texas.gov/?link=TN.

  “In Western culture…the investigator”: John E. Reid et al., Essentials of the Reid Technique: Criminal Investigation and Confessions (Sudbury, Mass.: Jones and Bartlett Publishers, 2005), p. 98.

  The Reid Manual is full of assertions about lie detection that are, to put it plainly, nonsense. The Reid “system” teaches interrogators, for example, to be alert to nonverbal cues, which have the effect of “amplifying” what a suspect says. By nonverbal cues, they mean posture and hand gestures and the like. As the manual states, on page 93, “hence the commonplace expressions, ‘actions speak louder than words’ and ‘look me straight in the eye if you’re telling the truth.’”

  If you stacked all the scientific papers refuting this claim on top of each other, they would reach the moon. Here is one of my favorite critiques, from Richard R. Johnson, a criminologist at the University of Toledo. (Johnson’s research can be found here: “Race and Police Reliance on Suspicious Non-Verbal Cues,” Policing: An International Journal of Police Strategies and Management 30, no. 2 [June 2007]: 277–90.)

  Johnson went back and looked at old episodes of the half-hour television documentary Cops. You may remember this show: it began in 1989 and still airs today, making it one of the longest-running programs on American television. A camera crew rides along with a police officer and films—cinema verité–style, without narration—whatever happens on that particular shift. (It’s strangely riveting, although it’s easy to forget that what you see on a typical Cops show is heavily edited; police officers simply aren’t that busy.) Johnson watched 480 old episodes of Cops. He was looking for interactions between a police officer and a citizen in which the citizen was on camera, from the waist up, for at least sixty seconds. He found 452 segments like that. Then he divided the segments into “innocent” and “suspect,” based on the information provided in the show. Was this the mother, child in arms, whose home had just been burglarized? Or was this the teenager who ran the instant he saw the police, and was found with the woman’s jewelry in his backpack? Then he subdivided his collection of clips one more time by race—white, black, and Hispanic.

  It should be pointed out that there is a small mountain of research on so-called demeanor cues. But Johnson’s study is special because it was not done in a college psychology lab. It’s real life.

  Let’s start with what many police officers believe to be the most important demeanor cue—eye contact. The Reid Technique’s training manual—the most widely used guide for law enforcement—is clear on this: People who are lying look away. Truthful suspects maintain eye contact.

  So what does Johnson find when he examines this idea in the light of real-world interactions on Cops? Are the innocent more likely to look an officer in the eye than the guilty?

  Johnson calculated the total number of seconds of eye contact per minute of footage.

  Black people who are perfectly innocent are actually less likely to look police in the eye than black people who are suspected of a crime. Now let’s look at white people:

  The first thing to note here is that Caucasians on Cops, as a group, look police officers in the eye far more than black people do. In fact, whites suspected of a crime spend the most time, of all four groups, looking the police officer in the eye. If you use gaze aversion as a cue to interpret someone’s credibility, you’re going to be a lot more suspicious of black people than white people. Far worse, you’re going to be most suspicious of all of perfectly innocent African Americans.

  OK. Let’s look at facial expressions. The Reid Technique teaches police officers that facial expressions can provide meaningful clues to a suspect’s inner state. Have I been found out? Am I about to be found out? As the manual states:

  “The mere fact of variation of expressions may be suggestive of untruthfulness, where the lack of such a variation may be suggestive of truthfulness” (Reid et al., Essentials of the Reid Technique, p. 99).

  This is a version of the common idea that when someone is guilty or being evasive, they smile a lot. Surveys of police officers show that people in law enforcement are very attuned to “frequent
smiling” as a sign that something is awry. To use the language of poker, it’s considered a “tell.” Here is Johnson’s Cops analysis of smiling. This time I’ve included Johnson’s data on Hispanics as well.

  Once again, the rule of thumb relied upon by many police officers has it exactly backward. The people who smile the most are innocent African Americans. The people who smile the least are Hispanic suspects. The only reasonable conclusion from that chart is that black people, when they are on Cops, smile a lot, white people smile a little bit less, and Hispanic people don’t smile much at all.

  Let’s do one more: halting speech. If someone is trying to explain themselves, and they keep nervously stopping and starting, we take that as a sign of evasion or deception. Right? So what does the Cops data say?

  The African American suspects speak fluidly. The innocent Hispanics are hemming and hawing nervously. If you do what the Reid manual says, you’ll lock up innocent Hispanics and be fooled by guilty African Americans.

  Does this mean we simply need a better, more specific set of interpretation rules for police officers? Watch out for the smooth-talking black guy. White people who don’t smile are up to no good. No! That doesn’t work either, because of the enormous variability Johnson uncovered.

  Take a look, for example, at the range of responses that make up those averages. Eye contact for innocent African Americans ranged from 7 seconds to 49.41 seconds. There are innocent black people who almost never make eye contact, and innocent black people who make lots of eye contact. The range for smiling for innocent black people is 0 to 13.34. There are innocent black people who smile a lot—13.34 times per minute. But there are also innocent black people who never smile. The “speech disturbances” range for innocent Caucasians is .64 to 9.68. There are white people who hem and haw like nervous teenagers, and white people who speak like Winston Churchill. The only real lesson is that people are all over the map when it comes to when and how much they smile, or look you in the eye, or how fluidly they talk. And to try to find any kind of pattern in that behavior is impossible.

  Wait! I forgot one of the Reid Technique’s big clues: watch the hands!

  During a response, a subject’s hands can do one of three things. They can remain uninvolved and unmoving, which can be a sign that the subject lacks confidence in his verbal response or is simply not talking about something perceived as very significant. The hands can move away from the body and gesture, which is called illustrating. Finally, the hands can come in contact with some part of the body, which is referred to as adaptor behavior. (Reid et al., p. 96).

  What follows is an explanation of how hand movements do and don’t contribute to our understanding of truthfulness. The Reid Technique assumes there is a pattern to hand movement. Really? Here are Johnson’s hand-movement data. This time I’ve included the range of responses—the shortest recorded response in the second column and the longest in the third column. Take a look:

  Hand gestures per minute

  Average time (in seconds)

  Shortest time (in seconds)

  Longest time (in seconds)

  African

  American/innocent

  28.39

  00.00

  58.46

  African American/suspect

  23.98

  00.00

  56.00

  Caucasian/innocent

  07.89

  00.00

  58.00

  Caucasian/suspect

  17.43

  31.00

  56.00

  Hispanic/innocent

  22.14

  23.00

  57.00

  Hispanic/suspect

  31.41

  13.43

  53.33

  Entire sample

  23.68

  00.00

  58.46

  If you can make sense of those numbers, you’re smarter than I am.

  By the way, the weirdest of all Reid obsessions is this: “Changes in [foot] bouncing behavior—whether it be a sudden start or stop—that occur in conjunction with a verbal response can be a significant indication of deception.…The feet are also involved in significant posture changes called ‘shifts in the chair.’ With this behavior, the subject plants his feet and literally pushes his body up, slightly off the chair to assume a new posture. Gross shifts in the chair of this nature are good indications of deception when they immediately precede or occur in conjunction with a subject’s verbal response” (Reid et al., Essentials of the Reid Technique, p. 98).

  What? I happen to be someone who is constantly, nervously jiggling his foot. I do it when I’m excited, or when I’m on a roll, or when I’m a little jumpy after too much coffee. What on earth does this have to do with whether or not I’m telling the truth?

  One more shot at the Reid Technique. Let me just quote from Brian Gallini’s devastating law-review article, “Police ‘Science’ in the Interrogation Room: Seventy Years of Pseudo-Psychological Interrogation Methods to Obtain Inadmissible Confessions,” Hastings Law Journal 61 (2010): 529. The passage is a description of a study done by Saul Kassin and Christina Fong: “‘I’m Innocent!’: Effects of Training on Judgments of Truth and Deception in the Interrogation Room,” Law and Human Behavior 23, no. 5 (October 1999): 499–516.

  More substantively, Professors Kassin and Fong videotaped one group of participants interrogated pursuant to the Reid method to determine whether they committed a mock crime. A second group of participants, some of whom were trained in the Reid method, watched the videos and opined on (1) the guilt or innocence of each subject, and (2) their confidence in their assessment of guilt or innocence. The results were as predictable as they were disturbing: First, judgment accuracy rates were comparable to chance. Second, “training in the use of verbal and nonverbal cues did not improve judgment accuracy.” In an effort to explain why training did nothing to improve judgment accuracy, the authors stated pointedly, “There is no solid empirical basis for the proposition that these same cues reliably discriminate between criminals and innocent persons accused of crimes they did not commit.”

  Finally, the authors reported, participants were overconfident in their assessment of guilt or innocence. In the authors’ words:

  [W]e found among both trained and naive participants that judgment accuracy and confidence were not significantly correlated, regardless of whether the measure of confidence was taken before, after, or during the task. Further demonstrating the meta-cognitive problems in this domain is that confidence ratings were positively correlated with the number of reasons (including Reid-based reasons) articulated as a basis for judgments, another dependent measure not predictive of accuracy. Training had a particularly adverse effect in this regard. Specifically, those who were trained compared to those in the naive condition were less accurate in their judgments of truth and deception. Yet they were more self-confident and more articulate about the reasons for their often erroneous judgments.

  “I apologize…these last couple of weeks…”: “Sandy Speaks—March 1, 2015,” YouTube, posted July 24, 2015, https://www.youtube.com/watch?v=WJw3_cvrcwE, accessed March 22, 2019.

  DOJ report on Ferguson, Missouri: United States Department of Justice Civil Rights Division, “Investigation of the Ferguson Police Department,” March 4, 2015, https://www.justice.gov/sites/default/files/opa/press-releases/attachments/2015/03/04/ferguson_police_department_report.pdf.

  African Americans are considerably more likely to be subjected to traffic stops (in footnote): Charles R. Epp, Steven Maynard-Moody, and Donald Haider-Markel, How Police Stops Define Race and Citizenship (Chicago: University of Chicago Press, 2004).

  North Carolina State Highway Patrol statistics: “Open Data Policing: North Carolina,” accessed March 2019, https://opendatapolicing.com/nc/, accessed March 2019.

  FM 1098 is not “a high-crime, high-drug area”: This crime map reflects Waller County data from 2013 to 2017 collected by Baltimore-based crime data aggregator SpotCrime, which sources data f
rom local police departments.

  More on the dilemmas caused by haystack searches: Middle-aged women, in most countries, are encouraged to get regular mammograms. But breast cancer is really rare. Just under 0.5 percent of women who get a mammogram actually have the disease. Looking for breast cancer is therefore a haystack search.

  Epidemiologist Joann Elmore recently calculated just what this means. Imagine, she said, that a group of radiologists gave a mammogram to 100,000 women. Statistically, there should be 480 cancers in that 100,000. How many will the radiologists find? 398. Believe me, for a task as difficult as reading a mammogram, that’s pretty good.

  But in the course of making those correct diagnoses, the radiologists will also run up 8,957 false positives. That’s how haystack searches work: if you want to find that rare gun in someone’s luggage, you’re going to end up flagging lots of hair dryers.

  Now suppose you want to do a better job of spotting cancers. Maybe getting 398 out of 480 cases isn’t good enough. Elmore did a second calculation, this time using a group of radiologists with an extra level of elite training. These physicians were very alert, and very suspicious—the medical equivalent of Brian Encinia. They correctly identified 422 of the 480 cases—much better! But how many false positives did that extra suspicion yield? 10,947. An extra two thousand perfectly healthy women were flagged for a disease they didn’t have, and potentially exposed to treatment they didn’t need. The highly trained radiologists were better at finding tumors not because they were more accurate. They were better because they were more suspicious. They saw cancer everywhere.

  If you are a woman, which group of radiologists would you rather have read your mammogram? Are you more concerned about the tiny chance that you’ll have a cancer that will be missed, or the much larger probability that you’ll be diagnosed with a cancer you don’t have? There’s no right or wrong answer to that question. Different people have different attitudes toward their own health, and toward risk. What’s crucial, though, is the lesson those numbers teach us about haystack searches. Looking for something rare comes with a price.